A State-Based Sensitivity Analysis for Distinguishing the Global Importance of Predictor Variables in Artificial Neural Networks

Table 5

Resultant mean values for input variables, to , and the modeled output, , within the case study simulation, #37 (see Figure 3), when values of lie within distinct “states” across its distribution (refer to Figure 2). : the instances of data vectors within a distinct “state” (total = 150).

	State of
2	−2.5	0.000	−0.313	−0.250	−0.313	−1.750
7	−1.75	−0.446	−0.161	−0.286	−0.268	−1.536
18	−1.25	0.118	−0.167	0.125	−0.375	−0.771
9	−0.875	−0.653	−0.083	−0.167	0.194	−1.111
10	−0.625	−0.300	−0.538	−0.050	−0.438	−0.575
18	−0.375	−0.007	−0.382	−0.458	0.111	−0.347
13	−0.125	−0.154	0.269	0.173	−0.115	−0.058
15	0.125	0.158	0.042	−0.383	−0.083	0.258
12	0.375	0.031	−0.031	0.406	−0.469	0.458
11	0.625	0.091	0.045	−0.250	0.011	0.420
8	0.875	0.281	0.297	0.578	0.344	0.797
17	1.25	0.096	0.272	0.125	0.434	0.838
6	1.75	0.313	0.500	0.021	0.313	1.500
4	2.5	0.344	0.844	0.656	1.281	1.688